DSLR Photography Forum

With much interest I noted Kombi making a correlation between high ISO and dropping sharpness, or no, you said the ability to focus precisely, right?

So funnily I think I ran right into that today. As I mentioned I am working on portraits and had a 200 picture session with this sweet old Italian Lady, to my regret,all indoors and at so so light.

inspired by Adrians fabulous waterpolo shots, I thought I would try to handle noise at higher ISO and shorter exp, instead of the other way round, and here is what happened.

When I looked at my pics , I was shocked that in almost none of the pictures her eye was well focused! I even started thinking something was wrong with my (kit-) lense.
I always pick one of the focus points and point it RIGHT at her eye, and I am quite positive about my aiming :-)

but then I also noticed, that the few shots in which I seemed to get it right, were the ones at the lowest ISO!




this is her at ISO 1600 and 1/100 sec.



a 100% crop of the wall, the noise level is horrific.



And her eye is so out of focus, that the shot is worth nothing.



now at ISO 200 and 1/13 sec (!!), a very similar shot looks like this:




the wall again just to show the - much more acceptable - noise.



and even though I find her eye is not as entirely in focus as I would have liked it to be, this is uncomparably better than at 1600.

FYI, shots at ISO 800 looked a little better, but similar to ISO 1600.



So is there really a correlation between noise and sharpeness / focus?
Actually, I should have said focus from the start, because I think nearby areas of the picture are very well in focus, but it is so off of the eye.....

Have you seen this before? How does it happen? and how do you get around it, if you have poor light and can't always use slooooooooow shutter speed?

uli

one more question for tonight, then I'll leave you alone :

I presume, a flash would help to some extend.

what kind of external flash does one use for portraits?

thanks everyone so much for advice and shoulders

uli

A couple of thoughts, without any real knowledge to back them up:

The eye is in shadow, so perhaps Canon's built-in noise reduction is reducing the detail; and

Is there a chance that the eye was too low contrast for the AF to lock accurately, causing backfocus? (No, I don't know why this might happen at high ISO, but not low ISO.)

I think Matthew is onto something with regard to the noise-reduction Uli.

I strongly suspect the camera (or RAW converter) really pumps up the noise-reduction to try to keep noise under control at 1600 ISO.. and the most common side-effect of any noise-reduction is a softening of the image.

While the 350D generally handles noise pretty well up to 800 iso (compared to other DSLRs of the same age and budget), there seems to me a huge difference between 800 and 1600. No wonder you can't push the 350D to 3200 iso like you can the 20D.

In addition, I am also becoming a bit less satisfied with the autofocus abilities of the 350D. People listed better AF as one of the benefits of a 20D over a 350D and I must admit I took that advice with a bit of a grain of salt at the time I bought my 350D... but now I get frustrated when the camera lets me down in situations where clearly it should have locked focus properly. I haven't had enough experience with a 20D to know for sure if it would fix these AF reliability concerns for me, but I suspect it would improve things.
The reason I mention this is because that might be contributing to the softness in some of these high-iso shots.
Think about when you are most like to use 1600 iso? In bad light.
And when is autofocus usually least reliable? You guessed it, in bad light.
So it makes sense that your camera is more likely to mis-focus in those same situations when you are more likely to be using 1600 iso.

Also its going to be tough to do these kind of sharpness tests on a human subject with shutter speeds of 1/13th of a second. Both you and the subject need to stay perfectly still, or you need to be using much faster shutter speeds.
Try the same shots outside in bright daylight. Use a fast shutter speed to eliminate camera shake and subject movement from introducing blur, and stop the aperture down a bit to minimise focusing complications... but don't stop the aperture down too much as it will start to soften the image again once it goes out of its "sweet spot" which is usually around f/8-f/11 for a lot of zooms.

I'd be interested to see how you go with this Uli.

I'm thinking about a test I saw, comparing colour rendition between a Canon 20D and E-1. It was looking at a textured brick wall under overcast conditions. It showed that the Canon gave a much sharper definition (strong local contrast) between the mortar lines and brick, but showed little colour variation within the reds of the brick. A sharp, smooth image. The Oly gave much lower tonal contrast, producing a less sharp look, but with a lot more per-pixel colour variation throughout the image. Olympus has a good reputation for rich colours and a bad reputation for noise, and that seems to be the source of both.

...that's why I was thinking that it was noise processing smoothing out the detail in the eye. There's plenty of sharpness in the stray hair, where it's very high contrast against the wall, but little per-pixel colour variation.

I think that's pretty much nailing it down, a little bit of poor focussing abilities, bad light,
and the noise reductions. That's such a smart thought, I would not have guessed. I wonder what the pictures would looke like without the noise reduction......

Unfortunately I can't always go outside so have to find a way to cope with indoor situations.

But it's true that the 350Ds autofocus has let me down sometimes before, and in these pictures it was clearly drawn to areas of higher contrast, despite the locked focus point.
In the manual it says you have 4 seconds to recompose after locking the focus, and I suspect that is not what's happening.